A job description and skills matrix you can steal.
Data teams are quietly rebuilding the same role under six different titles. Someone owns whether the metadata estate is complete enough for a dashboard, an agent, and an auditor to all get the same answer. That person usually has a different job title and no mandate.
This document gives that work a name, a scope, and a leveling path. Copy the description into your req. Copy the matrix into your promotion case. Nothing here is vendor-specific, and it will work against any metadata stack.
The role in one paragraph
A MetadataOps engineer runs the metadata estate as an operated system. They own the harvest, the modeling, the version history, the compilation into downstream tools, and the verification that all of it still matches reality. The work sits between data engineering, data governance, and AI platform engineering, and it borrows its operating discipline from all three: a build, a test, a deploy, a diff, and a rollback.
The shorthand: data engineers move the data. MetadataOps engineers own what it means, everywhere it lands.
Job description (paste-ready)
Summary
We're hiring a MetadataOps Engineer to own the completeness, accuracy, and governance state of our metadata estate. You'll be accountable for a measurable outcome: any consumer of our data, whether that's a BI report, an AI agent, or an auditor, gets the same definition of the same business concept.
This is an engineering role. The work happens in code, version control, and CI/CD.
What you'll own
Harvest. Coverage of the estate. You'll extend metadata collection across databases, transformation code, BI tools, and cloud platforms, and you'll know what percentage of the estate is covered and what's still dark.
Model. Business concepts bound to physical assets. You'll define metrics, dimensions, and vocabulary once, with an owner and a definition that survives contact with three BI tools.
Version. Change history for the metadata estate itself. Every definition change is logged, attributed, reversible, and answerable months later. You'll run dev, test, and production environments for metadata the way your platform team runs them for services.
Compile. Governed definitions written into the native formats of the systems that consume them, so downstream tools inherit the definition they consume.
Verify. Proof that the estate still matches reality. You'll build the checks that catch drift, run impact analysis before schema changes ship, and produce the evidence trail when someone asks where a number came from.
Responsibilities
- Own and report a metadata completeness score across five dimensions: coverage, resolution, derivation, freshness, and governance state.
- Build and maintain column-level lineage across the estate, and close the gaps where lineage is inferred from run logs.
- Define and govern business metric definitions with named owners, approval workflow, and version history.
- Run impact analysis ahead of schema and pipeline changes; publish the blast radius before the change lands.
- Deliver governed definitions to downstream consumers, including BI platforms and AI systems, through native formats or context protocols.
- Serve as the technical owner for audit and regulatory requests that ask how a number was produced.
- Partner with analytics engineering, data governance, and AI platform teams. Set the standards those teams work within.
Qualifications
Required
- Four or more years in data engineering or data platform work. Analytics engineering counts where it came with estate-wide scope.
- Strong SQL, plus one of Python, Scala, or Java.
- Hands-on with a transformation framework (dbt, Informatica PowerCenter, SSIS, Spark, or similar) and at least one BI semantic model (Power BI, Tableau, Looker, MicroStrategy).
- Working knowledge of lineage, cataloging, and governance concepts, and a clear view of where each one breaks down.
- Comfort across a heterogeneous estate, including systems that predate the current stack.
- Git-based workflow, CI/CD, and code review as normal practice.
Preferred
- Experience through a BI platform migration, on either side of it.
- Exposure to regulated reporting (SOX, BCBS 239, CSRD, ISSB) or to AI systems that query enterprise data.
- Familiarity with semantic modeling formats and context protocols such as MCP.
- Experience with legacy estate: mainframe, stored procedures, or on-prem warehouses alongside cloud.
Out of scope
Naming this matters, because the role gets diluted within a quarter otherwise:
- Data quality remediation. MetadataOps measures whether the metadata describing the estate is complete, current, and governed. Fixing bad values in the data itself belongs to data engineering and data quality tooling.
- Pipeline reliability and on-call. Different discipline, different pager.
- Manual glossary curation as a full-time activity. If the role becomes a curation queue, it has failed. The output is a system that stays accurate on its own.
- Prompt engineering and model tuning. MetadataOps supplies governed context. What an application does with it is the AI team's call.
Skills matrix
Levels are defined by work products. Whoever ships the artifact in the column holds the level.
DimensionMetadataOps EngineerSenior MetadataOps EngineerStaff / LeadHarvestExtends collection to new sources; maintains existing connectionsOwns coverage across a domain; can state coverage as a number with a denominatorSets the coverage target for the estate and the sequence for closing itModelDefines and documents metrics inside an existing standardDesigns the standard: naming, ownership, binding concepts to physical assetsOwns the ontology and resolves cross-domain conflicts where two teams claim the same wordVersionWorks comfortably in the metadata repo under version controlRuns dev, test, and production promotion for metadata changesDesigns the change management model, including approvals and rollback policyCompileShips governed definitions to one target platformShips to multiple targets and reconciles the differences between themSets the deployment architecture and decides what gets enforced whereVerifyRuns impact analysis on requestBuilds automated drift detection and publishes the resultsOwns the evidence model that satisfies audit, and can defend it in the roomInfluenceWorks within one teamSets standards other teams adoptCarries the business case to the CDO and gets headcount approved
Leveling notes
The clearest tell between Engineer and Senior is the denominator. An engineer can say lineage coverage improved. A senior can say it went from 61% to 84% of the estate and can show what's still dark.
The clearest tell between Senior and Staff is whether the standard survives without them in the room.
Making the case
If you're hiring: make it a rework argument. Count the hours your analytics team spends reconciling metric definitions across tools, then count the AI or BI initiatives currently waiting on a definition that has no owner. Both numbers are usually available, and both are usually worse than expected.
If you want this role: start by measuring. Score your estate on the five dimensions, write down what's dark, and take that document to your manager. A scored estate turns "we should own this" into "here's the gap, here's the quarter it closes, here's who owns it." That's the conversation that gets funded.
Written by the team at MetaKarta and released under CC BY 4.0: free to use, adapt, and republish with attribution. The role and the practice belong to the practitioners who named them.
